Clinical Pharmacokinetics and Pharmacodynamics of Nintedanib

Nintedanib (Vargatef^®, Ofev^®), also known as BIBF 1120, has the chemical name 1H-indole-6-carboxylic acid, 2,3-dihydro-3-[[[4-[methyl[(4-methyl-1-piperazinyl)acetyl]-amino]phenyl]amino]phenylmethylene]-2-oxo-,methyl ester, (3Z)-,ethanesulfonate and a molecular weight of 539.62 as free base. The chemical structure is shown in Fig. 1. Structural modelling studies with recombinant VEGFR-2 kinase demonstrated that nintedanib binds to the ATP-binding site in the cleft between the NH₂ and COOH terminal lobes of the kinase domain [14].

Nintedanib is dosed as a soft gelatine capsule. It displays a pH-dependent solubility profile with low aqueous solubility under neutral pH and increased solubility at acidic pH < 3 [24]. Nintedanib is considered to be a class 2 or 4 compound under the Biopharmaceutics Classification System (i.e. having low water solubility at neutral pH together with an ambiguous classification regarding permeability). Although nintedanib shows a high in vitro permeability, the absorption rate and bioavailability of nintedanib are decreased due to P-glycoprotein (P-gp) transporter effects [24].

The competitive binding of nintedanib to the ATP-binding pocket of the intracellular receptor kinase domain of VEGFR 1–3, PDGFR-α and -β and FGFR 1–3 inhibits the tyrosine kinase activity of these receptors, thereby interfering with the cross-autophosphorylation of the receptor homodimers and blocking the signalling cascade [25]. Nintedanib has been shown to inhibit the FGFR, PDGFR and VEGFR signalling cascades mediating lung fibroblast proliferation and migration [6, 16]. It inhibits protein kinase signalling pathways in three cell types contributing to angiogenesis, endothelial cells, pericytes and smooth muscle cells, resulting in inhibition of cell proliferation and apoptosis [14].

In cell-free in vitro kinase assays, nintedanib showed nanomolar potency to inhibit all three VEGFR subtypes, PDGFR-α and -β and FGFR-1, -2 and -3 [14]. In cell-based assays, inhibition of VEGFR-2 phosphorylation was sustained for at least 32 h in VEGFR-2-transfected NIH3T3 cells exposed to nintedanib for 1 h [14], suggesting a sustained inhibitory effect.

In preclinical disease models nintedanib showed antifibrotic effects in animal models of pulmonary fibrosis [16], highlighted by significant reductions in total lung collagen and reduced fibrosis in histological analyses. Nintedanib exerted anti-inflammatory effects as shown by significant reductions in lymphocyte and neutrophil counts in bronchoalveolar lavage fluid, reductions in inflammatory cytokines, and reduced inflammation and granuloma formation in histological analysis of lung tissue [16]. Additional data show that nintedanib significantly reduces lung collagen levels in a mouse model of rheumatoid arthritis-associated interstitial lung disease [26].

Nintedanib has also demonstrated antifibrotic effects in preclinical models of systemic sclerosis (SSc), an immune-mediated rheumatic disease of unknown aetiology that is characterised by fibrosis of the skin and internal organs and vasculopathy [27]. In two preclinical mouse models of SSc, nintedanib effectively inhibited fibroblast activation and exerts potent antifibrotic effects [28]. In a further mouse model, nintedanib ameliorated histological features of pulmonary arterial hypertension, destructive microangiopathy, and pulmonary and dermal fibrosis [29].

Nintedanib also effectively interfered with the formation and maintenance of the tumour vascular system resulting in tumour growth inhibition and stasis [14, 30]. Treatment with nintedanib in mouse xenograft models led to a rapid reduction in tumour microvessel density, pericytes vessel coverage and tumour perfusion [14] and resulted in significant tumour growth inhibition in models of different human cancer types including NSCLC [14].

For a more thorough discussion of the preclinical efficacy profile of nintedanib, the reader is referred to comprehensive reviews by Hilberg et al. [14] and Wollin et al. [6, 16].

The pharmacokinetics and drug metabolism of nintedanib (dosed via intravenous [IV] infusion or oral gavage) were studied in several animal species.

Mean plasma protein binding of nintedanib was > 97% in mice and rats, 91–93% in monkeys, and 98% in humans, over a concentration range of 50–2000 ng/mL [13, 17, 31]. Albumin was the major binding protein.

Following administration of [¹⁴C]-radiolabelled nintedanib to rats, radioactivity was widely distributed into most of the tissues (except the central nervous system [CNS]). Repeated oral dosing ([¹⁴C]-radiolabelled nintedanib 30 mg/kg) for 13 days showed a slight accumulation in some tissues, although a similar accumulation in plasma concentrations was not apparent [32].

Several in vitro metabolism, transport and drug interaction studies were performed to quantitatively assess the drug–drug interaction potential of nintedanib.

In vitro studies with human hepatocytes and/or human liver microsomes showed that nintedanib is a minor substrate for cytochrome P450 (CYP) 3A4 isoenzyme and has a very low potential (along with its two major metabolites [BIBF 1202 and BIBF 1202 glucuronide]) to inhibit or induce CYP isoenzymes [24], including those that are most relevant or genetically polymorphic for drug metabolism in humans (CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19 and CYP3A4) [32]. In human liver microsomes, nintedanib was rapidly hydrolysed by esterases (major mechanism) and demethylated by CYP3A4 (minor mechanism) to form the metabolites BIBF 1202 and BIBF 1053, respectively [33]. CYP-dependent metabolism accounted for about 5% compared to about 25% ester cleavage [20]. Thus, drug–drug interactions with nintedanib as a victim of CYP enzyme-modulating agents (e.g. with co-medication with CYP inhibitors or inducers) are considered very unlikely. Furthermore, drug–drug interactions with nintedanib as a perpetrator of CYP enzymes (e.g. nintedanib acting as a CYP enzyme inhibitor or inducer) are also considered very unlikely.

Further in vitro data indicated that nintedanib, at clinically relevant concentrations, did not inhibit glucuronidation by uridine 5′-diphospho-glucuronosyltransferase (UDP-glucuronosyltransferase, UGT) 1A1 (UGT1A1) in human liver microsomes. UGT1A1 is responsible for the glucuronidation of the metabolite BIBF 1202 to BIBF 1202 glucuronide in human liver microsomes [33]. In addition, BIBF 1202 was glucuronidated by several intestinal UGTs (UGT1A7, UGT1A8, UGT1A10). A clinically relevant drug–drug interaction based on inhibition of UGT after oral administration of nintedanib is considered less likely as all half-maximal inhibitory concentration (IC₅₀) values are substantially higher than the therapeutic plasma concentrations [32].

In vitro assays using transfected MDCK cells demonstrated that nintedanib is a substrate of the efflux transporter P-gp and weakly inhibits P-gp (Table 1) [24]. Studies using cell lines that express different drug transporters showed that nintedanib was not a substrate of organic anion-transporting polypeptide (OATP) 1B1, OATP1B3, OATP2B1, organic cation transporter (OCT) 2, multidrug resistance-associated protein 2 (MRP-2) or the efflux breast cancer resistance protein (BCRP), but was a weak substrate for OCT1 [31]. Nintedanib did not inhibit OATP1B1-, OATP1B3-, OATP2B1-, OCT1-, OCT2-, P-gp- or BRCP-mediated transport at clinically relevant concentrations [33].

Although the principal metabolite BIBF 1202 shows pharmacological activity at some of the target receptors, in vivo cellular activity indicates a substantially lower potency than nintedanib [32]. BIBF 1202 was not effective in mouse xenograft models [31], suggesting that BIBF 1202 plasma concentrations themselves are unlikely to contribute to the clinical effects of nintedanib in vivo. BIBF 1202 glucuronide did not show in vitro activity at the target receptors and is a non-reactive glucuronide [34].

In summary, in the clinical setting, nintedanib is unlikely to affect the pharmacokinetics of other drugs that are substrates of major drug-metabolising CYPs or uptake transporters, such as OATP1B1, OATP1B3, OAT1, OAT3 or OCT2. Drug–drug interactions with nintedanib as a victim of CYP enzyme-modulating agents (e.g. with co-medication with CYP inhibitors or inducers) are considered very unlikely. Nintedanib is a substrate of P-gp in vitro. The potential for a clinical drug–drug interaction between nintedanib and inhibitors or inducers of P-gp was addressed in dedicated phase I studies, as described in Sect. 5.

Following oral administration of soft capsules under fed conditions, nintedanib is absorbed promptly, with maximum concentration (C_max) being reached at 2–4 h (range 0.5–8 h) [25, 34, 35]. A representative plasma concentration–time profile of nintedanib after single 100 mg doses is shown in in Fig. 2. After reaching C_max nintedanib follows at least bi-phasic disposition kinetics.

The absolute bioavailability of the oral capsule formulation of nintedanib (100 mg) relative to IV dosing (4-h infusion, 6 mg) is around 5% (95% CI 3.6–6.1) [38]. Absorption and bioavailability are thought to be decreased by transporter effects and substantial first-pass metabolism [20].

After a high-fat, high-calorie meal [35] nintedanib exposure appears to be slightly increased, with C_max and area under the drug plasma concentration–time curve (AUC) from time zero to infinity (AUC_∞) increased by ~ 15% (geometric mean [gMean] ratio point estimate 115.3%, 90% CI 84.6–157.2) and 20% (gMean ratio 120.6%, 90% CI 95.3–152.5), respectively [24]. The 90% CIs of the AUC_∞ and C_max ratios under fed versus fasted conditions were not entirely within the range of 80–125% but include 100% in both cases. Absorption is delayed by about 2 h (from 2 to 4 h) in the fed state and inter-patient variability is increased. Nintedanib is recommended to be taken with food to improve gastrointestinal tolerability [19, 22].

Visual inspection of dose-normalised pharmacokinetic parameter estimates (C_max and AUC) show no relevant deviation from dose proportionality across the dose range of 50–450 mg once daily and 150–300 mg twice daily (Fig. 3) [38]. In line with this, the pharmacokinetics of nintedanib 150–250 mg twice daily in a PopPK analysis of patients with NSCLC or IPF were found to be dose proportional [49]. Clearance of nintedanib does not change after multiple dosing compared with single administration, meaning that the pharmacokinetics of nintedanib can be considered to be time independent. Thus, the pharmacokinetics of nintedanib are linear with respect to dose and time (i.e. single-dose data can be extrapolated to multiple-dose data) [20]. Based on trough concentrations, steady state is reached by 7 days at the latest (no sampling at earlier timepoints) [38, 42]. Accumulation after repeated twice-daily dosing is negligible (ratio of about 1.4- to 1.7-fold based on the AUC from time zero to 12 h [AUC₁₂]) [38] (Tables 2, 3). The pharmacokinetics of nintedanib do not differ substantially between healthy volunteers and patients with IPF or cancer.

Pharmacokinetic parameters show moderate to high inter-individual variability, as demonstrated by coefficients of variation (CVs) between 46 and 83% [25, 34, 44, 47] (see Table 2). This variability seems to be caused by differences in absorption rather than disposition and elimination. The low bioavailability (< 5%) and high first-pass metabolism of the drug may contribute to this observation [38]. Intra-individual variability is low to moderate (geometric CV [gCV] < 40%).

Nintedanib is ~ 98% plasma protein bound, with serum albumin considered to be the major binding protein [20]. After IV administration of nintedanib (6 mg), a high volume of distribution (gMean 1050 L, gCV 45%) is observed [38], indicating extensive distribution into peripheral tissues. This observation is in accordance with the high in vitro permeability of nintedanib through biomembranes in cellular assays and the results of a quantitative whole-body autoradiography study in rats, in which a rapid and relatively homogenous tissue distribution apart from the CNS and thus a pronounced blood–brain barrier was observed [17, 24]. Nintedanib is preferentially distributed in plasma with a blood to plasma ratio of 0.87 [20].

Nintedanib is predominantly metabolised via hydrolytic ester cleavage, with subsequent glucuronidation and excretion via the liver. The disposition and mass balance of nintedanib was investigated in a dedicated absorption, distribution, metabolism, and elimination (ADME) study in eight male subjects who received a single oral dose of 100 mg [¹⁴C-]-radiolabelled nintedanib solution [34]. The major route of elimination of drug-related radioactivity following a single oral dose of [¹⁴C]-radiolabelled nintedanib solution was via faecal/biliary excretion (93.4% in faeces within 120 h after dosing) with negligible clearance in the urine (0.65% of the dose within 72 h after dosing). Renal excretion of unchanged nintedanib was 0.05% within 48 h of dosing. The combined faecal and urinary recovery after 4 days accounted for 92.4% of the administered [¹⁴C]-radiolabelled dose, indicative of an essentially complete mass balance. Most of the recovery occurred within 24–48 h of dosing.

Metabolite profiling showed that metabolism is the major pathway of clearance for nintedanib, primarily through rapid hydrolytic ester cleavage of the methyl ester moiety, resulting in the formation of BIBF 1202, which is subsequently glucuronidated by UGT enzymes (UGT1A1, UGT1A7, UGT1A8 and UGT1A10) in the intestine and by UGT1A1 in the liver to form BIBF 1202 glucuronide [34] (Fig. 1). Most of the drug-related radioactivity determined in plasma was accounted for by nintedanib (24%), BIBF 1202 (32%) and BIBF 1202 glucuronide (30%). Other metabolites were only observed in low amounts in plasma or excreta. One of the minor metabolic pathways occurred via CYP enzymes, predominantly CYP3A4, and yielded a demethylated metabolite (designated BIBF 1053) that could not be detected in plasma but only in excreta, predominantly faeces (4% of dose) [19, 34].

These findings indicate that oxidative metabolism mediated via CYPs is of negligible importance for the metabolism and elimination of nintedanib, suggesting that the risk of drug–drug interactions between nintedanib and other therapies modulating CYP enzymes (i.e. CYP inhibitors and CYP inducers) is minimal.

In addition, nintedanib has no in vitro liability to inhibit or induce CYP enzymes (see Sect. 5), which may facilitate its combination with treatments for IPF (e.g. pirfenidone) and cytotoxic chemotherapies (e.g. docetaxel and paclitaxel) that utilise CYP enzymes.

The disposition and metabolism of nintedanib were also investigated following IV infusion (6 mg) [38]. Nintedanib showed a high total plasma clearance (gMean 1390 mL/min, gCV 28.8%), which classifies it as a high clearance drug. Comparison of exploratory metabolite pharmacokinetic data after IV and oral administration suggests that BIBF 1202 and its glucuronide are formed faster and to a larger extent after oral administration. The urinary excretion of unchanged drug within 48 h after IV administration was about 1% of dose, indicating that non-renal mechanisms are responsible for the clearance of nintedanib. The renal clearance was about 20 mL/min. Thus, renal excretion plays a minor role in the elimination process of nintedanib, both after IV and oral administration.

Elimination of nintedanib from plasma occurs with a terminal half-life of approximately 10–15 h (gCV approximately 50%) [20, 38]. The effective half-life of nintedanib in patients with IPF was estimated to be about 9.5 h (gCV 31.9%) based on PopPK model-predicted accumulation ratios (Table 3) [19].

Nintedanib is not being excreted via the kidneys to a relevant extent. The contribution of renal excretion after oral administration of nintedanib, both as unchanged drug (about 0.05% of dose) and as drug-related radioactivity (about 0.6% of dose), is minor.

In line with this, results from the PopPK analyses based on phase II and III data indicate that exposure to nintedanib was not influenced by mild (classified as creatinine clearance [CL_CR] 60–90 mL/min) or moderate (CL_CR 30–60 mL/min) renal impairment [49, 50]. Therefore, adjustment of the starting dose in patients with mild to moderate renal impairment is not required [19, 22] and no dedicated study in renally impaired patients has been performed. The safety and efficacy of nintedanib have not been thoroughly studied in patients with severe renal impairment (CL_CR < 30 mL/min) and therefore treatment in these patients is not recommended.

Administration of nintedanib in patients with moderate or severe hepatic impairment is not recommended, and patients with mild hepatic impairment should be monitored closely and the dose adjusted accordingly. Several studies have evaluated the effect of hepatic impairment on the pharmacokinetics of nintedanib. In a dedicated single-dose study of nintedanib 100 mg in subjects with mild (n = 8) or moderate (n = 8) hepatic impairment (Child-Pugh A and B grades [51], respectively) and matched healthy control subjects (n = 17) [39], exposure to nintedanib based on C_max and AUC_∞ was approximately two-fold higher in subjects with mild hepatic impairment and approximately eight-fold higher in subjects with moderate hepatic impairment than in matched controls (Table 4). Subjects with severe hepatic impairment (Child-Pugh C) have not been studied.

In addition, supportive data on the pharmacokinetics of nintedanib in patients with hepatic impairment have been collected as part of the clinical development programmes in IPF and cancer patients. Despite some differences in the classification of hepatic impairment (Child-Pugh classification not used), the pharmacokinetic data of nintedanib in individuals with hepatic impairment were aligned across datasets, with an increase in nintedanib exposure observed in patients with impaired liver function [39]. The supportive data comprise two phase I dose-escalation studies of open-label nintedanib in Asian and European patients with impaired hepatic function and advanced hepatocellular carcinoma (HCC) [40, 41] as well as the two PopPK analyses in patients with NSCLC and IPF [49, 50]. In both analyses, patients were defined as having mild hepatic impairment if their aspartate aminotransferase (AST) or alanine aminotransferase (ALT) or bilirubin levels were greater than the upper limit of normal (ULN), but AST and ALT were ≤ 10 × ULN and bilirubin was ≤ 1.5 × ULN at start of treatment. The number of patients with mild hepatic impairment at the start of treatment was 116 (of 849 patients with NSCLC and 342 patients with IPF) in the first analysis [49] and 44 (of 933 patients with IPF) in the second [50]. A trend toward elevated nintedanib exposure of up to 1.4-fold was observed in patients with mild hepatic impairment compared with patients with normal hepatic function. Because of missing information about underlying hepatic disease, a robust assessment of the effect of hepatic impairment defined by elevation of transaminases or bilirubin on nintedanib was not possible. Pharmacokinetic observations for nintedanib and its major metabolites in patients with hepatic impairment point towards an increase in the bioavailable fraction of nintedanib in these subjects (i.e. higher exposure in subjects with hepatic impairment than in healthy subjects, with comparable plasma concentration–time profiles). This is in line with nintedanib being a high clearance drug with high first-pass metabolism. Renal clearance of nintedanib is not influenced by impaired hepatic elimination.

Based on these findings, treatment with nintedanib is not recommended for patients with moderate or severe hepatic impairment (Child-Pugh B or C) [19, 20, 22]. Patients with mild hepatic impairment (Child-Pugh A) should be monitored closely and the dose adjusted accordingly. Prescribing guidelines in the USA and Europe recommend a dose reduction to 100 mg twice daily for patients with IPF and associated mild hepatic impairment [19, 20], while patients with NSCLC and associated mild hepatic impairment should be closely monitored and the dose adapted based on tolerability [22].

A drug–drug interaction study in healthy male subjects evaluated the effect of the potent P-gp inhibitor ketoconazole on nintedanib exposure [52]. Administration of ketoconazole increased exposure to nintedanib by 61% (AUC_∞) and 83% (C_max) (Fig. 4). Therefore, patients should be monitored closely for tolerability of nintedanib if a potent P-gp inhibitor (e.g. ketoconazole, erythromycin or cyclosporine [ciclosporin]) is being coadministered as exposure to nintedanib may be increased [19, 22]. Management of adverse effects may require interruption, dose reduction or discontinuation of nintedanib therapy.

In healthy male subjects [52], coadministration with the potent P-gp inducer rifampicin (rifampin) decreased plasma exposure of nintedanib to 50.3% (AUC_∞) and 60.3% (C_max) compared to administration of nintedanib alone (Fig. 4). Therefore, coadministration with P-gp inducers (e.g. rifampicin, carbamazepine, phenytoin and St. John’s wort) should be avoided (US label [19]) or carefully considered (EU label [22]) as this may result in decreased exposure to nintedanib.

The pharmacokinetics of nintedanib have been studied in combination with standard chemotherapy agents, including pemetrexed [53, 54], docetaxel [55, 56], afatinib [57, 58], carboplatin plus paclitaxel [59], cisplatin plus gemcitabine [60] and mFOLFOX-6 (leucovorin calcium [folinic acid], 5-fluorouracil, and oxaliplatin) [61]. None of the studies demonstrated a significant pharmacokinetic drug–drug interaction between nintedanib and the other therapies.

Pirfenidone is another medication approved for patients with IPF. A dedicated pharmacokinetic drug–drug interaction study investigated the combination of nintedanib and pirfenidone in two groups of patients with IPF using an intra-individual comparison of the pharmacokinetic data [62]. In a group of patients naïve to nintedanib or pirfenidone (n = 20), nintedanib plasma exposure was similar when coadministered with pirfenidone (titrated to 801 mg three times daily) in comparison to administration of nintedanib alone (Fig. 4). In a second group of patients treated with pirfenidone (n = 17), treatment with nintedanib (150 mg twice daily) for 7 days had no effect on the pharmacokinetics of pirfenidone. Further studies with this combination therapy in patients with IPF [63, 64] support the findings of this study that there is no evidence of a relevant pharmacokinetic drug–drug interaction between nintedanib and pirfenidone when administered in combination [20]. Only an exploratory safety study in Japanese patients with IPF showed a more pronounced trend towards decreased nintedanib exposure (by 32% for AUC and by 41% for C_max) [42], but these data should be treated with caution as exposure to nintedanib could only be compared between patient groups (inter-individually) and the number of patients included in the pharmacokinetic evaluation was low (< 12 per treatment group).

Nintedanib is also being investigated as a treatment for patients with SSc-associated interstitial lung disease (SSc-ILD) [65], which may be associated with pulmonary arterial hypertension. Bosentan, a dual endothelin receptor antagonist, is used in SSc-ILD patients to treat pulmonary hypertension and to prevent digital ulcers. As such, it might be frequently combined with nintedanib. Since bosentan is an inducer of CYP3A4 and CYP2C9, the potential for induction of P-gp can be assumed. The pharmacokinetics of nintedanib were studied in combination with bosentan [66]. In healthy subjects, pre-treatment with bosentan (125 mg twice daily) for 7 days had no effect on the single-dose pharmacokinetics of nintedanib 150 mg (Fig. 4).

In view of its mechanism of action (VEGFR inhibition), nintedanib has the potential for an increased risk of bleeding when used with medications that can alter haemostasis (e.g. antiplatelet therapy, anticoagulants). In the INPULSIS trials, low-dose antiplatelet therapy and prophylactic anticoagulation was permitted [8]. Patients at known risk for bleeding or treated with full-dose therapeutic anticoagulation were excluded. The frequency of patients who experienced bleeding was slightly higher with nintedanib (10.3%) than with placebo (7.8%); non-serious epistaxis and contusion represented the most frequent bleeding events [19, 20, 67]. Serious bleeding events occurred with low and similar frequencies in both groups (placebo group 1.4%, nintedanib group 1.3%). Based on these results, the prescribing information for nintedanib states that patients with a known risk of bleeding should be treated with nintedanib only if the anticipated benefit outweighs the potential risk, i.e. an individual risk assessment should be performed for each patient [19, 20]. Patients on full anticoagulation should be monitored for bleeding and have their anticoagulant treatment adjusted as necessary [19].

The effect of a single 200 mg dose of nintedanib and multiple 200 mg doses of nintedanib (administered twice daily for 15 days) on cardiac repolarisation was studied in 64 patients with advanced renal cell carcinoma (RCC) [47]. The upper limits of the two-sided 90% CIs for the adjusted mean time-matched QT interval using the Fridericia correction method (QTcF) from baseline to days 1 and 15 (primary electrocardiogram [ECG] endpoint) were well below the regulatory threshold of 10 ms at all times. The largest mean time-matched increase of QTcF at steady state was 3.1 ms (two-sided 90% CI − 0.2 to 6.4). No changes in the mean corrected QT (QTc) interval of > 60 ms were detected on days 1 or 15 and no patient had new-onset prolongation of the QTcF interval or an uncorrected QT interval of > 500 ms on days 1 or 15. There was no correlation between pharmacokinetic parameters and ECG variables and no treatment-related morphological ECG abnormalities were detected. Thus, the data indicate that nintedanib does not affect the QTc interval and, therefore, no specific ECG monitoring is indicated clinically.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is an established method for assessing the antivascular effects of antiangiogenic inhibitors [68]. Two studies used DCE-MRI to assess the size (bi-dimensional diameter) and vascular properties (vascular permeability, vascular density, regional flow) of suitable target lesions under nintedanib therapy in patients with advanced cancer [25, 48]. In both studies, changes consistent with an antivascular pharmacodynamic effect (decreased vascular permeability and tumour blood flow) were observed after nintedanib treatment. These included changes in key kinetic parameters such as K^trans (transfer constant), k_ep (reflux constant) or iAUC₆₀ (initial AUC for the first 60 s) of a gadolinium chelate complex. The effects were not clearly dose dependent, but most responses were seen at doses of ≥ 200 mg. Effects were already seen 24–48 h after the first drug intake and were preserved or even increased after continuous treatment with nintedanib over several weeks. DCE-MRI response was associated with disease stabilisation.

Exposure–response modelling of nintedanib in the treatment of patients with IPF was performed [69, 70] using data from 1403 patients who participated in the phase II TOMORROW and the phase III INPULSIS trials [8, 21] and received nintedanib doses of 50–150 mg twice daily or placebo. The association between nintedanib exposure (based on trough concentrations) and efficacy (based on the annual FVC decline in patients over time) and the relationship between nintedanib exposure and safety (based on the probability of experiencing diarrhoea or transaminase elevations ≥ 3 × ULN) was investigated.

The exposure–efficacy relationship was described by a linear disease progression model with a disease-modifying drug effect on the rate of FVC decline, indicating a slowdown of FVC decline with nintedanib treatment. A maximum effect (E_max) relationship between nintedanib exposure and FVC decline was established using observed and PopPK model predicted trough concentrations as exposure metrics. The trough concentration producing 80% of the E_max (EC₈₀) on FVC decline was estimated to be 10–13 ng/mL, which was similar to the median trough concentration achieved with the therapeutic dose of nintedanib 150 mg twice daily (10 ng/mL) during the phase III trials (Fig. 6). A stepwise covariate selection approach was applied to explore factors influencing the baseline FVC, natural disease progression or the drug-dependent treatment effect. None of the tested patient characteristics (age, height, sex, smoking status, Asian subpopulations [including Chinese, Taiwanese, Korean, Indian and Japanese ethnicity], presence of honeycombing, diarrhoea or FVC percentage predicted at baseline) was a distinct predictor of the treatment effect of nintedanib.

The safety data were described by parametric time-to-first event models [70] using the same exposure metrics as for exposure–efficacy correlations. A reliable association between nintedanib exposure and the risk of developing diarrhoea (any severity grade) could not be established, with nintedanib dose shown to be a better predictor for diarrhoea than exposure. This conclusion was based on several findings. First, an exposure–diarrhoea model (using a sigmoidal E_max relationship) was not superior to a model using dose as the predictor of diarrhoea risk. Second, visual predictive checks based on exposure–diarrhoea models and further exploratory analyses indicated that the diarrhoea risk for patients with the same exposure but receiving different doses was not comparable. Finally, results from covariate analyses suggested that patient populations with different exposure levels seemed to have a comparable diarrhoea risk. Therefore, no change in diarrhoea risk might be expected for patients with altered nintedanib exposure due to co-medications or individual patient characteristics. However, nintedanib exposure tended to be associated with a higher risk of developing liver enzyme elevations (defined as alanine transaminase and/or aspartate transaminase elevation ≥ 3 × ULN). The observed relationship was weak, and the analysis was limited by the low number of observed events (i.e. 38 safety events out of 895 nintedanib-treated patients). The effect of covariates on the relationship between exposure and liver enzyme elevations has not been investigated at present. Overall, the exposure–response analyses provide a modelling framework for a quantitative benefit–risk assessment in patients with IPF with altered nintedanib exposure due to co-medication or patient characteristics. The analyses support the 150 mg twice daily starting dose of nintedanib resulting in efficacious exposure levels in the majority of IPF patients (with the median exposure level just approaching the plateau of the established E_max relationship). Due to a potentially higher frequency of adverse events (e.g. in terms of liver enzyme elevations), close monitoring for tolerability is warranted for patients with elevated nintedanib exposure (e.g. due to Asian race, low body weight, older age or combinations of these risk factors).